Verdict: Over-annotated / weakly supported — the GO:0047400 annotation should be removed, or at minimum downgraded and qualified as in-vitro-only and non-core.
The single IDA annotation asserting that yjdM/P0AFJ1 has phosphonoacetate hydrolase activity (GO:0047400, EC 3.11.1.2) rests entirely on one machine-learning-prediction-plus-in-vitro study (DeepECtransformer, PMID:37963869(https://pubmed.ncbi.nlm.nih.gov/37963869/)). A subsequent peer-reviewed reassessment (PMID:40703034(https://pubmed.ncbi.nlm.nih.gov/40703034/)) explicitly concluded that "the PhnA activity observed in vitro is not supported," assigned the prediction the lowest possible confidence score, and called for additional in vivo experiments. Three independent lines of evidence — structural/evolutionary, genetic/historical, and genomic-context — converge to refute the homology-based premise of the annotation and to identify the underlying cause as a name collision between two unrelated genes both historically called "phnA."
YjdM is a small (111-amino-acid) dual zinc-ribbon protein (Pfam PF03831 + PF08274; InterPro IPR004624) whose four N-terminal cysteines form a single tetrahedral Cys4 zinc-binding site. The canonical phosphonoacetate hydrolase enzyme, PhnA from Pseudomonas fluorescens 23F, is a ~400-residue alkaline-phosphatase-superfamily binuclear-metal hydrolase that "showed no significant homology with any data bank accessions." These two proteins share neither size, fold, catalytic architecture, nor sequence. The E. coli "phnA" synonym for yjdM is a 1990 positional/alphabetical label for the first open reading frame in a sequenced phosphonate-region cluster — a gene that was subsequently shown by mutational analysis to lie outside the functional phnC–phnP phosphonate-utilization operon.
The most important caveat is that yjdM's genuine biochemical function remains experimentally uncharacterized. Refuting the phosphonoacetate-hydrolase assignment does not tell us what yjdM does; it only tells us what it almost certainly does not do. The better-supported molecular-function lead is zinc ion binding (via the Cys4 ribbon), and the gene should otherwise be treated as a conserved zinc-ribbon protein of unknown primary function.
The phosphonoacetate hydrolase (GO:0047400; EC 3.11.1.2) IDA annotation on yjdM/P0AFJ1 derives solely from the DeepECtransformer study (PMID:37963869(https://pubmed.ncbi.nlm.nih.gov/37963869/)), which combined a deep-learning EC-number prediction with a limited in vitro validation. That paper states: "We experimentally validated the enzymatic activities predicted for three proteins (YgfF, YciO, and YjdM)," making clear that YjdM's phosphonoacetate-hydrolase assignment is a machine-learning prediction with in-vitro follow-up, not an independently established biochemical characterization.
Critically, a peer-reviewed reassessment of exactly these predictions (PMID:40703034(https://pubmed.ncbi.nlm.nih.gov/40703034/)) reached the opposite conclusion. It states verbatim: "In conclusion, the PhnA activity observed in vitro is not supported, and additional in vivo experiments are required to confirm the biological role of this enzyme." This reassessment assigned the YjdM prediction the lowest confidence score in its scheme (CS = 1). When the sole source of an IDA annotation is directly contradicted by a later peer-reviewed paper, the IDA code — which signifies a direct, reproducible experimental demonstration of function — is no longer appropriate.
The same reassessment also undercut the genomic rationale for a phosphonate role: "However, except for E. coli, yjdM genes are generally not close to phosphonate catabolism or transport genes." In other words, the apparent association between yjdM and phosphonate metabolism is a coincidence of E. coli gene order and does not generalize across the ~12,000 members of the IPR004624 family.
The canonical phosphonoacetate hydrolase, encoded by phnA of Pseudomonas fluorescens 23F, is a novel carbon–phosphorus bond-cleaving enzyme of the alkaline-phosphatase superfamily. On its original cloning, "The nucleotide and deduced amino acid sequences of the phnA gene showed no significant homology with any data bank accessions" (PMID:9300819(https://pubmed.ncbi.nlm.nih.gov/9300819/)) — meaning it is not homologous to the small zinc-ribbon YjdM family. The true enzyme is ~400 residues and uses a binuclear metal center typical of its superfamily; its gene clusters with phosphonoacetate transport and LysR-type regulatory genes (phnR, phnB), not with zinc-ribbon genes (PMID:11344133(https://pubmed.ncbi.nlm.nih.gov/11344133/)).
By contrast, YjdM is a 111-residue dual zinc-ribbon protein (Pfam PF03831 "YjdM" + PF08274 "Zn_Ribbon_YjdM"; InterPro IPR004624), with CxxC cysteine pairs at C6/C9 and C23/C26. The AlphaFold model AF-P0AFJ1-F1 (mean pLDDT 78.2 over 111 residues; radius of gyration 22.2 Å) places the four N-terminal cysteines C6, C9, C23 and C26 in a tight tetrahedral cluster — all pairwise SG–SG distances between 3.4 and 3.8 Å and a centroid radius of ~2.2 Å — the exact geometry of a single Cys4 structural zinc site. The fifth cysteine (C93) sits 46–49 Å away and is not part of the metal site. This is a structural zinc-binding ribbon, not the α/β binuclear-metal active site of an alkaline-phosphatase-superfamily hydrolase. No shared fold, no shared active-site chemistry, no sequence homology: the homology premise behind the GO:0047400 transfer fails on every structural axis.
The association between yjdM and phosphonate metabolism is rooted in nomenclature, not biology. In 1990, Chen et al. (PMID:2155230(https://pubmed.ncbi.nlm.nih.gov/2155230/)) sequenced the E. coli B phosphonate region and identified "Seventeen open reading frames (phnA to phnQ) … in one transcriptional direction," naming them alphabetically by position. "phnA" was simply the first ORF (b4108/yjdM) in that stretch — a positional label, not a demonstrated enzyme, and entirely unrelated to the independently named Pseudomonas phosphonoacetate hydrolase gene phnA.
Metcalf and Wanner (PMID:8388873(https://pubmed.ncbi.nlm.nih.gov/8388873/)) then defined the functional operon by mutational analysis: "All genes for phosphonate (Pn) utilization in Escherichia coli are in a large cluster of 14 genes named, in alphabetical order, phnC to phnP." The functional phosphonate-utilization operon is phnC–phnP — it explicitly excludes phnA (yjdM) and phnB. Thus yjdM is genetically excluded from the C–P lyase pathway in its own organism.
Consistent with this, STRING (identifier 511145.b4108) returns the unknown-function neighbor yjdN as the top partner (score 0.917), followed by phn-operon genes (phnD 0.83, phnJ 0.82, phnH 0.82, phnG 0.79, phnK 0.76, phnI 0.72, phnC 0.71, phnM 0.71). These phn associations are driven by STRING's genomic-neighborhood channel and reflect the coincidental physical adjacency of yjdM to the phn operon in E. coli — precisely the confounder shown not to generalize across genomes (PMID:40703034(https://pubmed.ncbi.nlm.nih.gov/40703034/)). They are circular, not independent functional evidence. The UniProt record P0AFJ1 carries only the keyword "Hydrolase" and, notably, no zinc-binding site feature despite the clear Cys4 ribbon — confirming that the record's current functional identity is entirely downstream of the single disputed study.
The chain of reasoning that produced the GO:0047400 annotation, and where it breaks, can be summarized as follows:
Pseudomonas fluorescens 23F phnA E. coli b4108 "phnA" (= yjdM)
~400 aa, alkaline-phosphatase 111 aa, dual zinc-ribbon
superfamily, binuclear metal (PF03831 + PF08274 / IPR004624)
TRUE phosphonoacetate hydrolase Cys4 structural zinc site
(EC 3.11.1.2) (C6/C9/C23/C26, SG-SG 3.4-3.8 Å)
| |
| *** NAME COLLISION *** |
| both historically called "phnA" |
+----------------------+----------------------+
|
v
DeepECtransformer ML predicts EC 3.11.1.2 for YjdM
(PMID:37963869)
|
v
Limited in vitro assay -> IDA GO:0047400 on P0AFJ1
|
+---------------------+---------------------+
| |
v v
REFUTED by reassessment REFUTED by genetics/structure:
(PMID:40703034): - phnC-phnP operon excludes yjdM
"PhnA activity ... not (PMID:8388873; PMID:2155230)
supported" in vivo; CS=1 - no homology to true PhnA
yjdM not near phosphonate (PMID:9300819)
genes outside E. coli - AlphaFold: Cys4 Zn ribbon, not
a hydrolase active site
The evidence is best explained by a classic annotation-propagation failure: a lexical name collision between two phylogenetically and biochemically unrelated genes, amplified by (a) a machine-learning EC prediction, (b) a single in-vitro assay that a later study could not support in vivo, and (c) a genomic-neighborhood coincidence specific to E. coli. None of these constitute direct evidence that the YjdM gene product catalyzes C–P bond cleavage of phosphonoacetate.
The positive signal in the data points elsewhere. The conserved dual-zinc-ribbon architecture and the tetrahedral Cys4 geometry are the strongest, most reproducible structural features of YjdM. These support a zinc ion binding molecular function and suggest a nucleic-acid-associated or metal-structural role typical of zinc-ribbon proteins, though the specific biological process remains undetermined.
| Citation | Evidence type | Supports/Refutes/Qualifies | Claim tested | Key finding | Context | Confidence & limitations |
|---|---|---|---|---|---|---|
| PMID:37963869(https://pubmed.ncbi.nlm.nih.gov/37963869/) | Computational (ML) + in-vitro assay | Source of hypothesis (weak) | YjdM has phosphonoacetate hydrolase activity | DeepECtransformer predicted EC 3.11.1.2; "experimentally validated … for three proteins (YgfF, YciO, and YjdM)" | E. coli, in vitro | Single study; ML-driven; in-vitro only; not reproduced in vivo |
| PMID:40703034(https://pubmed.ncbi.nlm.nih.gov/40703034/) | Review/reassessment of primary predictions | Refutes | Whether YjdM PhnA activity is real | "the PhnA activity observed in vitro is not supported"; lowest confidence (CS=1); yjdM "generally not close to phosphonate catabolism or transport genes" outside E. coli | Cross-genome analysis | Peer-reviewed; directly addresses the exact prediction |
| PMID:9300819(https://pubmed.ncbi.nlm.nih.gov/9300819/) | Structural/evolutionary (sequence) | Refutes homology | Is YjdM homologous to true PhnA? | True phosphonoacetate hydrolase "showed no significant homology with any data bank accessions"; ~400-aa novel C–P lyase | Pseudomonas fluorescens 23F | Establishes the true enzyme family is sequence-distinct from YjdM |
| PMID:8388873(https://pubmed.ncbi.nlm.nih.gov/8388873/) | Mutant/genetic (TnphoA') | Refutes | Is yjdM part of the phosphonate operon? | Functional operon is "14 genes named … phnC to phnP" — excludes phnA(yjdM)/phnB | E. coli | Definitive genetic delimitation of the pathway |
| PMID:2155230(https://pubmed.ncbi.nlm.nih.gov/2155230/) | Sequence/historical | Refutes (name collision) | Origin of the "phnA" name | "Seventeen open reading frames (phnA to phnQ) … in one transcriptional direction" — alphabetical/positional naming | E. coli B | Shows "phnA" is a positional label, not an enzymatic assignment |
| PMID:11344133(https://pubmed.ncbi.nlm.nih.gov/11344133/) | Structural/functional (gene region) | Qualifies (true-enzyme context) | Genomic context of the real phnA | Real phnA clusters with phnR (LysR regulator) and phnB (transporter), not zinc-ribbon genes | Pseudomonas fluorescens 23F | Orientation for the authentic enzyme's pathway |
| AlphaFold AF-P0AFJ1-F1 | Structural (computational) | Refutes active site; supports Zn binding | Hydrolase active site or zinc site? | Cys C6/C9/C23/C26 form tetrahedral Cys4 cluster (SG–SG 3.4–3.8 Å); C93 is 46–49 Å away; mean pLDDT 78.2 | Model, 111 aa | Model-based; geometry unambiguous for a structural zinc ribbon |
| STRING 511145.b4108 | Interaction/association (database) | Competing artifact | Does yjdM functionally associate with phn genes? | Top partner yjdN (0.917); phn links (0.71–0.83) via neighborhood channel | E. coli | Neighborhood-channel-driven; circular with E. coli gene order |
Lead requiring curator verification: Remove, or downgrade and qualify, the GO:0047400 (phosphonoacetate hydrolase activity) IDA annotation on P0AFJ1.
The net recommendation is that GO:0047400 should not be treated as yjdM's function, and yjdM should be curated as a conserved zinc-ribbon protein of unknown primary function with a zinc-binding structural feature.
The immediate molecular function being tested is the direct catalysis of C–P bond cleavage of phosphonoacetate (EC 3.11.1.2) by the YjdM polypeptide. The evidence relevant to that direct gene-product activity is: (a) one ML prediction + in-vitro assay that was not reproduced in vivo; (b) the absence of any sequence or structural homology to the true enzyme family; and (c) a structural model showing a zinc-binding ribbon rather than a hydrolase active site. None of these support direct catalytic activity.
The signals that might superficially suggest a phosphonate role are all downstream or contextual, not direct: the "phnA" name (a lexical artifact), the E. coli-specific genomic neighborhood (a gene-order coincidence), and STRING associations (a neighborhood-channel inference derived from that same coincidence). No loss-of-function phenotype, no in vivo growth-on-phosphonate dependency, and no biochemical reconstitution establishes a phosphonate-catabolic role for yjdM.
Name collision (primary alternative explanation). Two unrelated genes are both called "phnA": the Pseudomonas fluorescens 23F phosphonoacetate hydrolase (PMID:9300819(https://pubmed.ncbi.nlm.nih.gov/9300819/); PMID:11344133(https://pubmed.ncbi.nlm.nih.gov/11344133/)) and the E. coli b4108/yjdM ORF (PMID:2155230(https://pubmed.ncbi.nlm.nih.gov/2155230/)). The GO transfer appears to have flowed across this lexical bridge rather than across genuine homology.
Organism-specific genomic context. The yjdM–phn adjacency is specific to E. coli; across the broader IPR004624 family it does not hold (PMID:40703034(https://pubmed.ncbi.nlm.nih.gov/40703034/)). Any functional inference from E. coli gene order is therefore not generalizable.
In-vitro-only artifact. The reassessment (PMID:40703034(https://pubmed.ncbi.nlm.nih.gov/40703034/)) frames the in-vitro activity as unsupported in vivo — consistent with a spurious or promiscuous low-level activity under non-physiological conditions, a well-known risk when validating ML enzyme predictions.
Database carry-over. The UniProt "Hydrolase" keyword and the GO IDA together now form a self-reinforcing record even though they derive from a single disputed study; the missing zinc-binding feature annotation shows the record under-represents the best structural evidence.
Alternative interpretation favored by the data: YjdM is a zinc-ribbon protein whose conserved function involves zinc binding and, plausibly, nucleic-acid association (a common role for zinc ribbons), but whose specific biological process is uncharacterized.
| PMID | Title (abbrev.) | Role in this assessment |
|---|---|---|
| 37963869 | DeepECtransformer ML enzyme prediction + validation | Sole source of the GO:0047400 IDA (ML + in-vitro) |
| 40703034 | Reassessment of the predictions | Refutes: in-vitro PhnA activity "not supported" in vivo; CS=1 |
| 9300819 | Cloning of Pseudomonas phosphonoacetate hydrolase | True enzyme is non-homologous to YjdM |
| 11344133 | phnA gene region analysis in P. fluorescens 23F | True enzyme's pathway/regulatory context (phnR, phnB) |
| 8388873 | TnphoA' mutational analysis of the phn operon | Functional operon is phnC–phnP; excludes yjdM |
| 2155230 | Sequence of the E. coli phosphonate region | "phnA" is a positional/alphabetical label (name collision) |
| 21261917 | Phosphonoacetate biosensor via PhnR | Orientation on the true phosphonoacetate degradative system |
The literature cleanly separates into two groups. The Pseudomonas phosphonoacetate-hydrolase papers (PMID:9300819(https://pubmed.ncbi.nlm.nih.gov/9300819/), PMID:11344133(https://pubmed.ncbi.nlm.nih.gov/11344133/), PMID:21261917(https://pubmed.ncbi.nlm.nih.gov/21261917/)) describe the authentic EC 3.11.1.2 enzyme — large, non-homologous to YjdM, and embedded in its own phnR/phnB regulatory and transport context. The E. coli genetics papers (PMID:2155230(https://pubmed.ncbi.nlm.nih.gov/2155230/), PMID:8388873(https://pubmed.ncbi.nlm.nih.gov/8388873/)) show that the E. coli "phnA" label is positional and that yjdM sits outside the functional phnC–phnP operon. The two modern papers (PMID:37963869(https://pubmed.ncbi.nlm.nih.gov/37963869/) and PMID:40703034(https://pubmed.ncbi.nlm.nih.gov/40703034/)) are, respectively, the origin and the refutation of the annotation under review.
Candidate action change:
- GO:0047400 (phosphonoacetate hydrolase activity): remove or downgrade/qualify — change evidence away from IDA; mark in-vitro-only, non-core; add curator note citing the refuting reassessment.
Candidate new/replacement GO terms:
- Add GO:0008270 zinc ion binding (MF) as a structural/computational lead (Cys4 ribbon; IPR004624; AlphaFold tetrahedral geometry) — label as ISS/structural, not IDA.
- Retain GO:0005829 cytosol (CC) if independently supported.
- Do not add any phosphonate catabolic process (BP).
Candidate references and exact snippets to verify:
- PMID:40703034(https://pubmed.ncbi.nlm.nih.gov/40703034/): "In conclusion, the PhnA activity observed in vitro is not supported, and additional in vivo experiments are required to confirm the biological role of this enzyme." and "However, except for E. coli, yjdM genes are generally not close to phosphonate catabolism or transport genes."
- PMID:37963869(https://pubmed.ncbi.nlm.nih.gov/37963869/): "We experimentally validated the enzymatic activities predicted for three proteins (YgfF, YciO, and YjdM)."
- PMID:8388873(https://pubmed.ncbi.nlm.nih.gov/8388873/): "All genes for phosphonate (Pn) utilization in Escherichia coli are in a large cluster of 14 genes named, in alphabetical order, phnC to phnP."
- PMID:2155230(https://pubmed.ncbi.nlm.nih.gov/2155230/): "Seventeen open reading frames (phnA to phnQ) were identified in one transcriptional direction."
- PMID:9300819(https://pubmed.ncbi.nlm.nih.gov/9300819/): "The nucleotide and deduced amino acid sequences of the phnA gene showed no significant homology with any data bank accessions."
Suggested curator questions:
- Does the GO IDA annotation cite any evidence independent of PMID:37963869? (Finding: no.)
- Has the in-vitro activity ever been reproduced in vivo? (Finding: no — explicitly unsupported per PMID:40703034.)
- Should UniProt P0AFJ1 gain a zinc-binding site feature (C6/C9/C23/C26)? (Recommended: yes, pending confirmation.)
Suggested experiments: the five discriminating tests above, prioritizing the in vivo phosphonoacetate growth assay in a yjdM deletion background.
Prepared for AI Gene Review hypothesis-level curation. All quantitative and structural claims above are drawn from the cited primary literature, the AlphaFold AF-P0AFJ1-F1 model, and public STRING/UniProt records. Computational results (AlphaFold geometry, STRING scores) are reported as model/database-derived inferences, distinct from direct experimental results.